Research On Decoding Algorithm And Performance Analysis Of Turbo Codes

Turbo codes are a class of high-performance error correction codes in which Convolutional codes and a random interleaver structure are combined to form approximate random long codes. Turbo codes use soft output iterative decoding to approach the maximum likelihood decoding and come closest to approaching the Shannon limit. Turbo codes become very potential for a lot of communication systems because of its excellent performance, especially when they are used within low SNR (signal to Noise Ratio) channels. And they have already been the standard for channel coding in the third and the forth generation mobile communication systems. It promotes the research and application of turbo codes greatly.After an introduction to the coding and decoding principles of Turbo Codes and analysis of their performance, the thesis is dedicated to the research on the decoding algorithm and the iterative stopping criterion. Firstly, the effects of a few design parameters on the BER performance of Turbo codes such as the component coding, the length of the interleaver, the number of iteration and coding rate are simulated and discussed. Accordingly, a basic principle of parameter selection for design of Turbo codes is provided. Secondly, the capacity and complexity of two kinds of decoding algorithm, MAP and SOVA(Soft Output Viterbi Algorithm), are simulated and analysized. There are some defects in traditional SOVA, such as bad performance, high decoding sequence delay and high storage requirement. Aiming to eliminate these defects, a new algorithm named SW-MISOA (Slide-Window Modified Soft-Output Viterbi Algorithm) is presented. On one hand, a correction function is introduced to correct SOVA soft output values exceeding predefined threshold values. As a result, the performance of SOVA is improved. On the other hand, the application of slide windows reduces the effects of decoding sequence delay and storage requirement. Lastly, common stopping criteria of iterative decoding are compared and discussed, then a new stopping criterion is brought up based on reliability following principles of Cross Entropy and SOVA decoding algorithm. The corresponding simulation and analysis shows that the new criterion does help reduce the average number of decoding iteration and the time delay while the BER performance is almost not degraded. Moreover, the complexity of computation and the storage memory is reduced, which is helpful for hardware implementations.